HylaFAX The world's most advanced open source fax server

[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [hylafax-users] Possible race condition



Hello (& thanks for the reply),

We are running version 4.4.4 so I'm really happy to hear about the optimizations:)

In regard to your question about the faxes being send concurrently;  The first fax transmission starts as soon as it has been enqueued but the subsequent 29 are all marked as blocked and the job sleeps until its time to send (TTS) has been reached.  The result in other words is that fax 1 is send straight away and faxes 2 through 30 are sent after 3 minutes (or whatever the default retry time is).

Hope this helps?
Dominique

On Mon, Oct 27, 2008 at 8:42 PM, Aidan Van Dyk <aidan@xxxxxxxx> wrote:
* Dominique dHotman <dominique.dhotman@xxxxxxxxxx> [081027 14:19]:
> Hello List,

> When I run this script the 1st fax transmission is started but the
> subsequent 29 are marked as blocked.  The following is logged:
>
>
>
> Oct 24 07:20:16 hylafax01 FaxQueuer[4357]: NOTIFY: bin/notify
> "sendq/q429" "blocked" ""

But they *are* sent concurrently, right?

In 4.4, the new queueing mechanism is optimized for large queues, with large
numbers of jobs, including large numbers of jobs to the same destination, in
sorted lists, based on future send times and destinations.

A "blocked" job is any job that's behind a job to the same destination.  What'
happening is that ifyou submit jobs fast enough, the previously job to that
dest hasn't actually been started yet, so the job is "behind another job to
that destination" (i.e. blocked).  But when faxq actually get's around to
starting jobs on devices, it knows about batching, and concurrent calls, and
can quickly grab them all (up to max concurrent calls, of course, sorted based
on total job queue priority) and start them.

When  you "pause" between sending, faxq actually starts the job on a device
before the next one, so as you submit them, they are never "behind" another job
to that destination.

> The reason that I'm worried about this is that we are going to be
> operating with fairly high volumes and blocked faxes where lines are
> available will cause a major major head ache :-(

Well, if you have huge queues, then you definately want 4.4!


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iD8DBQFJBguruVxNPsxNPScRAvMMAKCw43FkxA3R+xGRDq1MrfPOsKgTWgCgoOC5
0XoPrk4L2LGHPgOj5kIyNVk=
=0jhj
-----END PGP SIGNATURE-----





Project hosted by iFAX Solutions